Power Imbalance and the Dark Side of the Captive Agri-food Supplier–Buyer Relationship

No business relationships are absolutely light or dark, but instead embody a mix of these two (Abosag et al., 2016). Previous literature (e.g. Håkansson and Snehota, 1998) emphasizes the dark side as deriving from the value of relationships. The literature (e.g. Abosag et al., 2016) has identified various situations that constitute the dark side, such as relationship unrest (Good & Evans, 2001), relationship burdens (Håkansson and Snehota, 1998), relationship stress (Holmlund-Rytkönen & Strandvik, 2005), the adverse sides of business relationships (Strandvik & Holmlund, 2008), relational misconduct (Hawkins et al., 2008), detrimental intentions (Liu et al., 2014), and exploitative business relationships (Schleper et al., 2017).

The concept of a ‘dark side’ to business relationships suggests “‘problems’, ‘challenges’, ‘difficulties’, and ‘drawbacks’ related to structural issues … such as size differences, or the imbalance of power; processes within business relationships, including creativity issues, capability development, changes in market dynamics; and outputs, for example, performance, competitiveness and satisfaction” (Abosag et al., 2016, p. 5). The dark side relates to negative business situations (e.g. Gaski, 1984; John, 1984). Regarding collaboration in service networks, the dark side refers to those aspects of co-creation activities that are hidden and include potential risks in business interactions (Chowdhury et al., 2016). Resource integration activities lead to value co-creation; however, particular relevance is the co-destruction of value which aligns with the dark side of business relationships. This assertion comes from the recognition that engaging in value co-creation may not always lead to symmetrical value outcome or desirable outcome (Dong et al., 2008; Chowdhury et al., 2016). Value can be co-destroyed when resources are accidentally or intentionally misused (Plé & Cáceres, 2010). Moving beyond power and dependence in dyadic relationships to a focus on triads and the wider network in supply chains, we argue that cooperative and joint actions have the potential of limiting the undesirable effects of the dark side. This is because even close dyadic relationships that seem to be the most stable are vulnerable to the dark side phenomenon (Anderson & Jap, 2005). The light side implies that positive benefits and desirable outcomes are derived from the relationship; it denotes the situation in which good relationship quality yields achievable relationship functions and performance (Fang et al., 2011).

Power is the ability of one channel member (A) to get another channel member (B) to do something that it otherwise would not have done (Coughlan et al., 2006; Dahl, 1957; Gaski, 1984). It refers to the capability of a partner in an exchange to induce another partner to change its behaviour in favour of the objectives of the partner exerting the influence (Wilemon, 1972). The association between power and dependence is better understood from Emerson's (1962) theory. We use ‘dependence’ and ‘dependency’ interchangeably with the same meaning and intended purpose. Dependency exists between two channel members when the benefits derived from their relationship are not available outside it. Thus, A’s power over B increases with B’s dependence on A. In imbalanced power relationships, the weaker firm is highly dependent on a more powerful firm or its agent to continue to achieve its goals. Whenever one partner (i.e. the principal) depends on another (i.e. the agent) to undertake some action on the principal’s behalf, this leads to a principal–agent relationship (Eisenhardt, 1989; Fontrodona & Sison, 2006). Power and dependence in exchange relationships lead to vulnerability, and the possibility of one partner or its agent being opportunistic and unethical. Dependence is a function of the greater utility (i.e. value, benefits, satisfaction) that B gets from A and the fewer alternative sources of that utility that B can find. It has to do with how easily B can replace A. The dependence condition creates a relational trap in which the weak partner has no choice but to endure the opportunism of the other (Grandinetti, 2017).

To varying extents, firms always depend on their trading partner. Early studies on dependence focused on the effects for the buyer of its dependence on the supplier, without considering the supplier’s dependency (e.g. El-Ansary & Stern, 1972). A few studies have incorporated dependence from the perspective of both the buyer and the supplier (Buchanan, 1992; Geyskens et al., 1996; Kumar et al., 1995). Dependence is defined, in psychology and social psychology, as a state in which assistance from others in terms of finance, emotion, protection, security or daily care, is expected or actively sought (Zhou et al., 2007). In the context of buyer–seller relationships and channel studies, dependence refers to the extent to which a trade partner provides important and critical resources for which there are few alternative sources of supply (Buchanan, 1992), or “the degree to which the target firm needs to maintain its relationship with the source in order to achieve its desired goals” (Kale, 1986, p. 390).

The context is the cocoa industry of Ghana. Certain cocoa-producing countries have a fully liberated local market with a free-market system characterized by many private exporters. While in others (e.g. Ghana) private, former state marketing monopolies retain substantial control and play ‘coordinative’ role in chain governance (Glavee-Geo, 2019). In Ghana, the industry is partially liberated, characterized by the participation of private firms and many cocoa growers as the main suppliers. The industry regulator is the Ghana Cocoa Board (COCOBOD).

We collected primary data from the cocoa growers, who were very knowledgeable about the issues at stake. The respondents were either farm owners or farm managers serving as key informants. We based the sampling on farms located in the southern part of Ghana according to the knowledge of the industry regulator. Approval was sought from each informant before each interview. Subsequently, primary data were collected through face-to-face interviews. In most developing and some emerging countries, data collection through mail or email leads to low response rates—hence the need for some other innovative means of data collection through household interviews. The lead author conducted the interviews for study 1 over a period of two weeks and for study 2 a year later for a period of five weeks in the cocoa-growing regions of Central, Eastern, and Ashanti regions of Ghana. The respondents were mostly farm owners who were interviewed in their houses (sometimes with a farm manager providing corroborative information). Study 1 consisted of 105 respondents while study 2 consisted of 444 respondents.

All constructs are based on reflective multi-item scales adapted from previously validated scales (see Tables 1 and 2). The indicators are measured on a seven-point Likert scale, with 1 representing the lowest level (strongly disagree) and 7 the highest level (strongly agree). Subjective measures of supplier financial performance are newly formulated, while the profitability measure is adapted from Haugland et al. (2007). The difficulty in accessing financial data makes the use of subjective measures of performance a better proxy to evaluate the performance of a supplying firm. The three-item scale of supplier financial performance is formulated with the anchors 1, representing worse performance, and 7, representing better performance. Single-item indicators of the relative size of the buyer, relationship duration, transaction/sales volume, supplier dependency, switching costs and buyer control of price were formulated based on the extant literature (e.g. Buvik & Andersen, 2015; Caniels & Gelderman, 2007). A single item—‘this buying company ensures that the minimum producer price does not vary’—measured buyer control of price. Supplier dependency was measured with the question ‘how large do you perceive your dependency on this particular buying company compared to other buying companies within this district?’ Switching costs were measured with ‘how much will it cost you if you want to replace this buying company with another one in a new location?’ The quality standardization item was formulated based on buyer control measures by Buvik and Andersen (2015). Power imbalance items were adapted from Joshi and Stump (1999). In study 1, power imbalance items were operationalized based on two main questions: ‘how large were your sales to this buying company?’ and ‘how large do you perceive the cash bonuses paid by this buying company to your farm business?’ However, in study 2, we sought to improve measurement of the power imbalance through reformulation of the two additional questions from study 1. Reliability results of studies 1 and 2 show the power imbalance scales had high internal consistency and extracted more than 50% of the variance in each instance (see Tables 1 and 2). Membership of cooperatives in study 2 was measured with the question ‘are you a member of any cooperative farmers association?’ with responses coded as a dummy, 1 for ‘yes’ and 0 for ‘no’. Opportunism items were adapted from John (1984) and Skarmeas et al. (2002). Five items were used in study 1, while seven items were used in study 2 with the same intended purpose of achieving good internal consistency. Economic satisfaction items were adapted from Geyskens and Steenkamp (2000) and Skinner et al. (1992), social satisfaction items from Benton and Maloni (2005) and Crosby et al. (1990).

We conducted the estimation using the partial least squares (PLS) structural equation modelling technique SmartPLS 3.0 (Ringle et al., 2015). PLS’ strength lies in its ability to deal with complex models with a high number of constructs, indicators and relationships (Hair et al., 2017). It is less strict with assumptions about the distribution of the data and equally ideal for small sample size (Chin & Newsted, 1999; Hair et al., 2017). The use of categorical variables with unknown non-normal frequency distribution, which are usually negatively skewed, makes PLS preferable. The preceding factors make it an ideal analytical technique for the current study and hence a preferable alternative to the use of maximum likelihood methods.

The descriptive statistics of the variables (mean and standard deviation), factor loadings, reliability, and average variance extracted (AVE) of the constructs for study 1 and 2 are presented in Tables 1 and 2.

The standardized loadings for the indicators and bootstrap t-values for all the items used in both study 1 and study 2 (see Tables 1 and 2) were all significant at 0.001 in the two-tailed test, except the item ‘The buying company often undervalues my cocoa beans’, which is significant at p < 0.01 (Table 1). All Cronbach’s alphas exceed the 0.7 threshold (Nunnally, 1978), while composite reliabilities (Fornell & Larcker, 1981) are higher than 0.8, showing high internal consistency. The average variance extracted (AVE) (Fornell & Larcker, 1981) exceeds 0.50, the lowest being 0.52 for the construct buyer/agent opportunism in study 1 and the highest 0.82 for the constructs supplier financial performance (study 2) and power imbalance (study 1) (see Tables 1 and 2). A higher AVE indicates that the variance captured by each latent variable is significantly larger than the variance due to measurement error, demonstrating unidimensionality and high convergent validity of the constructs.

We assessed discriminant validity (Chin, 1998; Fornell & Larcker, 1981) by comparing the square root of the AVE for each construct with the correlations of all other constructs in the model (see Tables 3 and 4). A correlation between constructs exceeding the square roots of their AVE indicates that they may not be sufficiently discriminable (Coelho & Henseler, 2012; Hair et al., 2017). A comparison of the square root of the AVE (bold face diagonal values) and the correlations among the constructs shows that the square roots of AVE are always higher than the absolute correlations between the constructs. We also checked the heterotrait–monotrait ratio of correlations and found all the values were under 0.85, demonstrating high discriminant validity (Henseler et al., 2015). We conclude that the measurement models for study 1 and study 2 show evidence of acceptable validity.

Common method bias (CMV) is variance attributable to the measurement method rather than to the constructs, this is because the data for all the model variables came from the same respondents at the same time. CMV might influence some of the hypothesized relations in the structural model (Podsakoff et al., 2003; Podsakoff & Organ, 1986). To avoid bias from CMV, we applied an a priori method (Hulland et al., 2018). This was done by careful design of the questionnaire and formulation of the question instruments. During the data collection, we conducted the interviews for both studies by asking the questions in a random order: this ensured that the dependent and independent variables in the survey were separated. Before the administration of both surveys, we pre-tested the questionnaire. The pre-test helped us avoid ambiguous question items that could be difficult to understand or interpret. Taking these steps prior to administering the survey helped to limit the potential for CMV.

The structural model results of study 1 (n = 105) and study 2 (n = 444) were estimated using SmartPLS 3 (Ringle et al., 2015). To evaluate the structural models of both studies, we first assessed the structural models for collinearity (Hair et al., 2017) by examining the variance inflation factor (VIF) values of all the predictor constructs. We found all the VIF values to be below the threshold of 3.3 (Diamantopoulos & Siguaw, 2006). We concluded that collinearity was not at critical levels (Table 5). Thereafter, we examined the significance and relevance of the path coefficients based on the results of the bootstrapping procedure with 10,000 subsamples (Franke & Sarstedt, 2019).

Though a substantial number of studies have looked at the role of power and dependency in business exchanges and relationships, the role that power and dependency play in the dark side phenomenon is less understood. This implies that the mechanisms of the dark side phenomenon are not fully explained, and therefore our knowledge of these negative business practices is still insufficient. Our study presents several contributions to fill the knowledge gap. The study provides empirical evidence of the moderating role of transaction volume between supplier dependency and power imbalance consistent with the literature (Anderson & Weitz, 1989; Geyskens et al., 1996). Besides, our findings show that asymmetric power relations have a negative impact on the gains of the other party in the exchange and this is further exacerbated by dependency. This is particularly true when there are few alternatives for the dependent partner, such that the relationship becomes a captive one and prone to exploitation, consistent with the literature.

The effect of switching costs in such captive relationships is equally acknowledged in the literature, highly cited as the cause and most often assumed to be a linear relationship. Our contribution to the literature shows switching costs impact in imbalanced business relationships to be curvilinear. Similarly, the association between power imbalance and opportunism has been much cited (e.g. Grandinetti, 2017; Hawkins et al., 2008; Tangpong et al., 2015), but with little evidence on whether this association is linear or curvilinear. Our study reaffirms the linear association and shows that the effect of power imbalance on opportunism could also be curvilinear.

The study also provides empirical evidence to show the consequences of the dark side phenomenon. Power imbalance is associated with reduced financial performance of the weaker partner (Chicksand, 2015; Griffith et al., 2006; Hingley, 2005a). The concept of collective action refers to an action taken by a group of people with the aim to achieve a common goal and to improve their social or economic situation. Collection action is differentiated by a group/joint action and decision instead of that of an individual. Cooperative supply and marketing associations are typical examples. Our findings show significant differences between cooperative and non-cooperative groups in relation to the impact of power imbalance on financial performance. The theoretical implication is that the negative impact of the dark side phenomenon could be averted to a greater extent with group/joint action and concerted effort.

Based on the research context, the cocoa agri-food supply market, the suppliers are heavily dependent on buyers and their agents. The impact of dependency on the power imbalance was found to be contingent on transaction volume, while the effect of power imbalance on financial performance was found to be moderated by supplier dependency. The implication is that small suppliers’ participation in collective action through supply and marketing cooperatives can help offset the asymmetrical power relations and improve individual and group welfare (Fischer & Qaim, 2012, 2014). Power can be used to control rewards to disadvantage the other party in the exchange; however, group action can provide a powerful voice for such disappointing incidents, relationship problems, and problematic business relationships. Collective action is a constructive group effort to change objectionable relationship conditions with the intent to improve conditions.

We draw on Dabholkar et al. (1994, p. 133), 'the interactions through which the terms of exchange are worked out are often referred to as the negotiation process'. Using the theory of negotiation (bargaining) behaviour (Clopton, 1984; Dabholkar et al., 1994; Perdue et al. 1986; Pruitt, 1981), negotiation behaviour can be classified along two dimensions of ''time'' and ''gain'' perspectives. Four categories of negotiation behaviour are identified: competitive, command, coordinative and cooperative. However, the terms ''cooperative'' and ''coordinative'' are used interchangeably in the literature. Competitive negotiation behaviour emphasizes individual short-term gain where parties maximize their outcomes. Command behaviour seeks out to maximize individual gain; however, the specific strategies are less 'coercive' than competitive bargaining behaviour.

Coordinative strategies focus on the long-term joint gain, while the cooperative approach is characterized by the short-term joint gain (Dabholkar et al., 1994). Though the above-negotiating strategies apply mostly to dyadic exchanges, it can also apply to network relations although more complex. Exchange relations between cocoa suppliers and lead firms can be classified along the two dimensions of short/long-term and individual/joint gain. Individual cocoa farmers dealing with lead firms and the government parastatal organization (Ghana Cocoa Board) have less bargaining power than if organized as cooperatives.

An important strategy for protecting firms from unethical behaviour is information sharing (Eckerd & Hill, 2012). The sharing of records and accounts to reduce the information asymmetry between buyer and supplier firm can help stimulate trust. This action is a critical obligation as emerging markets liberalize their agri-food supply markets and exporters compete for market share for cocoa and coffee in the originating countries (Dadzie et al., 2018). Others have recommended accredited ethical purchasing agents (Buxton & Vorley, 2012), where lead firms dealing with small suppliers can educate, train and align the objectives, incentives and motivations of their sourcing managers, frontline staff and intermediaries to be ethical with suppliers (Vorley & Thorpe, 2014). Building ethical and mutually beneficial relationships between small dependent suppliers and lead firms is an important strategy to reduce the negative consequences of exploitative business relationships and to control opportunism (Glavee-Geo et al., 2020).

Corporate ethical values (CEV) and formal ethical infrastructure are key features of an organization that can be relied on to ensure employees' attitudinal and behavioural change. This calls for clear ethically responsible purchasing practices, codes of conduct, and enforceable company policies (Saini, 2010). However, ethical codes are not enough. Hyman et al. (1990) suggests the need for a checklist for evaluating managerial decisions, and to improve the chances of being ethical. CEVs represent the amount of attention afforded to ethical issues by the firm, and the degree to which the firm behaves ethically (Hunt et al., 1989). A firm's formal ethical infrastructure (FEI) is characterized by formal communication, recurrent communication, formal surveillance, and formal sanctions and has been found to influence employee’s moral awareness (Hawkins et al., 2013; Rottig et al., 2011). The ethical infrastructure of an organization includes a range of organizational systems that guide ethical decision-making (Hawkins et al., 2013). Hence the communication of ethical standards to employees engenders moral behaviour (Eckerd and Hill, 2012).

Laudable as some of the strategies may be, more so when most of the initiatives are at the lead firm's benevolence, the challenge is how can resistant lead firms be convinced? Global supply chains have become overly complex and more vulnerable to disruptions with large unanticipated consequences (Fahimnia et al., 2015). Also, notwithstanding major disruptions either human-made or natural (e.g. COVID-19) and supply risk caused by several sources of inherent uncertainties such as demand fluctuations, supply capacity changes, lead time variability, and exchange rate volatility (Fahimnia et al., 2015), firms face enormous pressure and responsibility to deliver consistently. The need to ensure consistent supply and be resilient implies that firms have no choice than to be ‘compliant’, socially responsible and enforce ethical behaviour among employees (sourcing and purchasing agents, intermediaries, and staff).

Supplier scarcity has been cited (Schiele et al., 2012) as an important reason for supplier resource mobilization and supplier development. Firms compete not only on the sales market but also on the supply market since ‘really good’ suppliers are scarce (Cordón & Vollmann, 2008; Schiele et al., 2012, p. 1178). Suppliers can actively differentiate among competing buyers based on their ethical standards such that ethical behaviour can be a differentiating attribute and cue. Thus, corporate ethical values can be used as a ‘discriminating’ signal to communicate a firm’s attractiveness to suppliers and beat competing buyers (especially regarding strategic products/services). ‘Since more and more sales revenue comes from purchased items, it becomes more critical to get the best—ahead of your competitors’ (Cordón & Vollmann, 2008, p. 14). Being ethical is being ‘smart’. The behaviours and activities (regarding ethicality) of the buying company become a key means of influencing suppliers' resource mobilization. It is also a means of communicating the firm's corporate reputation to other stakeholders (government agencies, customers, civil society organizations, and the wider public) and reinforcing the firm's social responsibility. Besides, unethical opportunistic behaviour can also be a costly problem within the organization and lead to substantial inefficiencies and reputational challenges.

The dark side phenomenon leads to challenges that inhibit creativity, innovation, and value creation (Abosag et al., 2016, p. 5). Value formation is not only associated with value co-creation but also with value co-destruction (Echeverri & Skålén, 2011). The co-destruction of value is one of the downsides of value co-creation, and this has ethical implications for business relations. Besides, power imbalance impedes supply chain innovation and improvement by the less dominant partner. Such a power imbalance can lead to the use of coercive power by the more dominant buyer. Small suppliers must devote more resources to ensuring the continuation of business relationships with powerful buyers at the expense of supply chain innovations (Matanda et al., 2016). Our study suggests that buyer control of price is associated with reduced financial performance, while buyer control of quality in terms of quality standardization is associated with higher financial performance. In addition, we found that decision control regarding quality standardization can enhance the coordinative role of power in improving performance. The implication is that the use of power for coordination can be more effective if this is done within an environment characterized by mutuality, honesty, empathy, and fairness (Woiceshyn, 2011).

Finally, the positive impact of economic satisfaction on financial performance was found to be stronger for non-cooperative members than for cooperative members. This could be explained by the fact that cooperative members expect more gains from their exchanges than non-cooperative members. Cooperative members invest more time and effort in collective action. The implication is that although collective action may not solve all challenges emanating from problematic business relationships, joint/collective action is one of the most viable options in opposition to the negative impact of the dark side. For example, fair trade initiatives that strive for more equitable sharing of profits among the members are typical examples of sustainable collaborative supply chains (Drake and Schlachter, 2008).